Lawyers are supposed to cite real cases. When they use AI tools to help write legal briefs, they sometimes cite cases that never existed.
According to a report by i24NEWS, attorneys have submitted court filings containing fabricated legal precedents, an outcome produced when AI language models generate false information with the same confident tone they use for accurate information. The phenomenon is widely called hallucination, a term for when AI systems produce outputs that are plausible-sounding but factually wrong.
The problem has moved from isolated incidents to a pattern visible enough to attract attention from courts, bar associations, and legal observers. In some cases, the fabricated citations appeared in filings submitted to real judges, who then had to address them. Courts have responded with sanctions, warnings, and new disclosure requirements asking attorneys to confirm whether AI was used in drafting documents.
The situation is complicated further when clients themselves take actions that damage their own cases, either by using AI tools without their attorney's knowledge, providing AI-generated research to their lawyers as if it were reliable, or making decisions based on AI-generated legal advice they found outside of their counsel's guidance. The i24NEWS report describes clients sabotaging their own cases as a distinct and growing category of the problem.
AI tools are not designed to fabricate, but the way large language models work makes hallucinations a structural risk rather than a bug that can be simply patched. These models predict what text should come next based on patterns in training data. When asked about a specific court case, a model may generate a case name, citation, judge, and ruling that all sound real because they follow the patterns of real legal writing, even when no such case exists.
The legal profession has particular exposure to this problem because legal briefs depend entirely on accurate citation. A fabricated precedent is not just wrong; it can mislead a judge, waste court time, and in some cases result in sanctions against the attorney who submitted it. The American Bar Association and several state bar associations have issued guidance on AI use in legal practice, though standards vary by jurisdiction and enforcement remains uneven.
The cases documented in the i24NEWS report span multiple courts and practice areas. The common thread is attorneys or clients relying on AI output without verifying citations against actual legal databases. Services like Westlaw and LexisNexis can confirm whether a cited case exists, a step that takes minutes but has been skipped with damaging results.
As AI tools become cheaper and more embedded in everyday legal work, the pressure to use them grows. So does the risk of the errors they produce reaching a courtroom before anyone checks.
